Journal of Entomology and Zoology Studies 2015; 3(4): 91-97 E-ISSN: 2320-7078 P-ISSN: 2349-6800 Multi-criteria geographic distribution approach: JEZS 2015; 3(4): 91-97 © 2015 JEZS A case study of Rhyssomatus nigerrimus Fhåraeus Received: 14-06-2015 Accepted: 15-07-2015 (Coleoptera: Curculionidae) Jose Lopez-Collado Colegio de Postgraduados, Jose Lopez-Collado, Catalino Jorge Lopez-Collado Campus Veracruz, Km 88.5 Carretera Xalapa-Veracruz, Veracruz, Mexico, Phone and Abstract Fax: 011 52 (229) 2010770. The objective of this study was to estimate the potential distribution of Rhyssomatus nigerrimus in Mexico. Point presence data and bioclimatic data were used to construct the model. Kernel density Catalino Jorge Lopez-Collado distributions were estimated for the bioclimatic layers to represent the response function and to combine Colegio de Postgraduados, them. The vulnerability of municipalities planted with soybean was computed as a function of suitability Campus Veracruz, Km 88.5 and revenue. Results showed a medium to high potential occurrence in the northern states of Mexico: Carretera Xalapa-Veracruz, Tamaulipas, North of Veracruz in the Gulf of Mexico and Sonora and Sinaloa in the Pacific Ocean. Low Veracruz, Mexico, Phone and suitability occurred in the northern and central areas of the country. The Area under the Curve was 0.805 Fax: 011 52 (229) 2010770. and the True Skill Statistics was 0.625, both indices supporting the validity of the model. Soybean- planted municipalities in Tamaulipas, San Luis Potosi, Chiapas and Campeche were estimated to have a medium to high vulnerability to pest occurrence. Keywords: Soybean pests, analytic hierarchic process, species distribution modeling, Curculionidae. 1. Introduction The Mexican Soybean Weevil (Rhyssomatus nigerrimus) Fhåraeus (Coleoptera: Curculionidae) is native to America and is found in Mexico, Barbados, Martinique, St. Vincent, Belize, Guatemala, Panama and South America [1]. It was first reported in Mexico from collection trips in the states of Veracruz and Guerrero [2]. Later, it was recorded in Guanajuato in wild hosts and in Nayarit and Tabasco [3, 4]. Adults of the Mexican Soybean Weevil have different host plants, including Acacia spp., Amaranthus spp., Ficus spp., and [4] Ipomoea spp. Recently, it was found as a pest of soybeans in San Luis Potosi, Tamaulipas and Chiapas [5, 6]. In soybeans, it feeds on the vegetative and reproductive structures. Adults feed on plants and seedlings, attacking stems, branches, sprouts, and buds. The larvae feed inside the pods; later, it drops to the soil, transforms to pupa and overwinters as adult. The insect shows one generation per year [5]. Soybean is planted in Mexico in about 178,530 ha in the states of Campeche, Chiapas, Chihuahua, Jalisco, Nuevo Leon, Quintana Roo, San Luis Potosi, Sonora, Tamaulipas, Veracruz and Yucatan, generating revenue near US$ 110 million [7]. The geographic distribution of organisms is a key factor in risk analysis and pest control. Habitat distribution is estimated by means of correlative, deductive and hybrid methods. Inductive or correlative methods rely on georeferenced point data, collected by sampling or by [8] specimens stored in museums and bioclimatic layers . Deductive methods use the physiological response of organisms to bioclimatic variables such as temperature and precipitation, which are known to influence the geographic distribution of species; hybrid techniques combine both methods [9]. [10] The diverse results obtained from different modeling tools suggest combining them . Alternatively, input layers can be combined based on their similarity to conditions where the species are known to thrive, using ad-hoc procedures [9]. This approach requires processing Correspondence: different data layers, which are combined by a simple or weighted sum [11, 12]. The analytic Jose Lopez-Collado hierarchic process AHP combines data from different sources in such a way that it uses Colegio de Postgraduados, [12, 13] Campus Veracruz, Km 88.5 heuristics to rank the relative importance of the data by making pairwise comparisons . Carretera Xalapa-Veracruz, The ranks are converted to weights and assigned to the data layers which are grouped in Veracruz, Mexico, Phone and categories for combination. This method has been applied to estimate suitable land areas for Fax: 011 52 (229) 2010770. rice production [14], determine the potential distribution of forest pests [15] and to select landfill ~ 91 ~ Journal of Entomology and Zoology Studies zones [16]. Because R. nigerrimus is a pest in its initial layer respectively, while j refers to any cell value of the i-th colonization stage of soybean crops in Mexico, it is important raster layer. This procedure was applied at each step depicted to know its potential geographic range and which in Fig. 1. The suitability map was aggregated using the k-mean municipalities with soybeans may be affected. Therefore the clustering algorithm to obtain four classes indicating different goal of this study was to apply the AHP to find the potential degrees of geographic suitability: unsuitable, low, medium and distribution of R. nigerrimus and estimate soybean regions high suitability. susceptible to potential damage in Mexico. 2. Materials and Methods This work applied a composite hierarchic approach to compute the potential distribution of R. nigerrimus [17], and it was realized during 2013-2015 from data compiled at different dates. The first step was extracting the bioclimatic data layer values from the points where the specimens have been reported, these values were then transformed using a response function. In the second step, weights associated to each layer were computed based on pairwise comparisons; weights corresponded to the normalized eigenvalues of a pairwise comparison matrix representing the variables grouped by categories. Weights derived from the pairwise comparisons allow ranking the relative importance of each layer. The third Fig 1: Hierarchic approach applied to compute the distribution model step involved combining the layers by using map algebra for R. nigerrimus. The X (i)'s are the input variables or data layers, operations. The model was evaluated by splitting the the Xc (i)'s are the composite layers; the w's are weights used for occurrence data in two sets: one used for training and another combining the layers to obtain Yf, the final distribution map. for testing. Crop vulnerability was computed by multiplying the potential distribution index by revenue. 2.4 Model evaluation Model performance was evaluated by using the independent 2.1. Specimen data test data to estimate two different metrics: the Receiver To estimate the potential distribution of R. nigerrimus, geo- Operating Characteristic (ROC) curve and the corresponding referenced point data were pooled from collection reports [2, 4, Area under the Curve (AUC), and by computing the True Skill 6], the data set was split in n=25 (70 %) for training and n=10 Statistics (TSS). These metrics were computed using the test (30 %) for testing. data and 10,000 background point data extracted randomly from the suitability map [22]. The AUC is a threshold 2.2. Selection of bioclimatic layer values and response independent index that tends to one when the model is function appropriate and to 0.5 when it cannot be distinguished from a [18] Values were extracted from numeric and nominal random model. The TSS range from -1 to 1, the closer to 1 the bioclimatic layers at the points where the species was present. better the model [23]. The TSS requires converting the The numeric variables were: mean annual temperature, mean suitability map to a presence (1)-absence (0) map U by temperature of coldest quarter and mean annual precipitation. applying a threshold value th, so that U= 1 if Yf th, 0 Nominal values for soil and climate types were also extracted otherwise. TSS value was computed from observed data point [19, 20] . These variables are thought to affect some biological at the lower suitability value. For the AUC, test data was [6] traits of the species, such as overwintering and survival . To resampled to obtain an estimate of the confidence interval [24]. combine the layers it is necessary to standardize the raster [12] values between cero and one , cero meaning no suitability 2.5 Vulnerability of soybean crop and one indicates maximum suitability. Different The vulnerability of soybean to this species was estimated as transformation functions have been proposed depending of the the product of mean potential occurrence and total revenue at problem domain. For numeric variables in the sample, the the municipality level [11], using data for year 2013 [7], revenue response was modeled by a kernel density distribution, while values were normalized between zero and one [12]. Operations for categorical variables each value was ranked according to were performed with R 3.1.2 and Mathematica 8.01 [25, 26]. its frequency to obtain a discrete density estimate. The layers were then transformed to their corresponding density 3. Results and Discussion estimates. In both cases, the kernel distributions reflected the 3.1. Kernel distributions and layer combination information contained in the observed values so that the most The kernel density estimates are presented in Fig. 2 and 3 for frequent received the highest response. the numeric and nominal sample points. The estimates represented the expected suitability values for the bioclimatic 2.3. Assessment and pooling of bioclimatic layers layers
Details
-
File Typepdf
-
Upload Time-
-
Content LanguagesEnglish
-
Upload UserAnonymous/Not logged-in
-
File Pages7 Page
-
File Size-